This disclosure describes a method to minimize disruption for throughput oriented jobs in power oversubscription services with a dynamic control. The mechanism controls power in a hardware-agnostic way, and the policy employs a multi-threshold approach that balances power safety with workload impact. Moreover, an alternative control mechanism ensures proper system operation while power measurements are unavailable.

Example methods and apparatus to facilitate dynamic core selection are disclosed. An example apparatus includes a first processor core of a first type; a second processor core of a second type different from the first type; and software to: access a user-supplied hint indicative of a user preference to execute program code on the first processor core, the user-supplied hint including a user-defined attribute of the program code; monitor performance of the program code on the first processor core; determine, based on the user-defined attribute of the program code, a predicted performance of the program code on the second processor core is better than the performance of the program code on the first processor core; and ignore the user preference by migrating the program code from the first processor core for execution on the second processor core

A virtual machine placement system for placing a plurality of virtual machines on a first physical server and a second physical server in order to efficiently operate a physical server in which the plurality of virtual machines are installed is disclosed. The physical server includes the first physical server and the second physical server, the virtual machine placement system contains a workload calculation module, a prediction module, a temperature prediction module, a schedule module, and a migration module, wherein the schedule module calculates a placement schedule considering predicted temperature of the physical server.

A method of managing computational and power resources in a datacenter includes receiving an application request at an allocator to execute a requested application, identifying an idle computing device in the datacenter, obtaining an efficiency parameter for the idle computing device, obtaining a normalized power demand of the requested application, and determining a device power demand for the requested application on the idle computing device based at least partially on the efficiency parameter and a normalized power demand for the requested application.

A system and method for determining optimal computing configuration for executing a computing operation includes defining one or more constrains of a given computing operation to be executed. The method further includes implementing a knowledge graph to determine at least one suitable combination of computing hardware and computing software based on the given computing operation and the defined one or more constrains therefor. The method further includes quantitatively estimating an energy requirement and qualitatively estimating an energy consumption pattern of the determined at least one suitable combination. The method further includes performing a life cycle assessment for execution of the given computing operation utilizing the determined at least one suitable combination of computing hardware and computing software based on the quantitative estimation of the energy requirement and the qualitative estimation of the energy consumption pattern, to determine an optimal combination of computing hardware and computing software therefor.

Embodiments of systems and methods for power throttling of High Performance Computing (HPC) components are described. In some embodiments, an HPC platform may include: a system Baseboard Management Controller (BMC), and an accelerator tray comprising a tray BMC coupled to a plurality of managed subsystems and to the system BMC, where the system BMC is configured to: in response to a power excursion event, instruct the tray BMC to throttle a first managed subsystem by a first amount and to throttle a second managed subsystem by a second amount.

Disclosed is a task allocation method, apparatus, electronic device, and computer-readable storage medium. The task allocation method includes: in response to receiving a synchronization signal, executing, by the master processing core, a task update instruction to obtain a to-be-executed task segment; receiving, by a processing core for executing the task, the to-be-executed task segment, wherein the processing core for executing the task includes the master processing core and/or the slave processing core; executing, by the processing core for executing the task, the to-be-executed task segment; and in response to completion of execution of the to-be-executed task segment, sending, by the processing core for executing the task, a synchronization request signal, wherein the synchronization request signal is configured to trigger generation of the synchronization signal.

Technologies for power-aware scheduling include a computing device that receives network packets. The computing device classifies the network packets by priority level and then assigns each network packet to a performance group bin. The packets are assigned based on priority level and other performance criteria. The computing device schedules the network packets assigned to each performance group for processing by a processing engine such as a processor core. Network packets assigned to performance groups having a high priority level are scheduled for processing by processing engines with a high performance level. The computing device may select performance levels for processing engines based on processing workload of the network packets. The computing device may control the performance level of the processing engines, for example by controlling the frequency of processor cores. The processing workload may include packet encryption. Other embodiments are described and claimed.

Examples relate to an apparatus, a device, a method, and a computer program for controlling the execution of a computer program by a computer system comprising two or more different Processing Units (XPUs), and to a corresponding computer system. The apparatus comprises processing circuitry configured to obtain the computer program, wherein at least a portion of the computer program is based on one or more compute kernels to be executed by the two or more different XPUs. The processing circuitry is configured to determine, for each XPU, an energy-related metric for executing the one or more compute kernels on the respective XPU. The processing circuitry is configured to assign the execution of the one or more compute kernels to the two or more different XPUs based on the respective energy-related metric.

Task scheduling in a computing device may be based in part on voltage regulator efficiency. For an additional task to be scheduled, multiple task scheduling cases may be determined that represent execution of the additional task on each of a number of processors concurrently with one or more other tasks executing among the processors. For each task scheduling case, a regulator input power level for a voltage regulator may be determined based on a performance level indication associated with the additional task, the one or more other tasks executing on the processors, and the efficiency level of each voltage regulator. For each task scheduling case, a total regulator input power level may be determined by summing the regulator input power levels for all voltage regulators. The additional task may be executed on a processor associated with a task scheduling case for which total regulator input power is lowest.